OpenAI Launches Frontier, an Enterprise Platform for AI Agents

OpenAI has launched Frontier, a platform for building, deploying, and managing enterprise AI agents, with State Farm as an early customer and availability starting today for a limited set of buyers. The company says broader availability will follow over the next few months.

The announcement, made in a blog post titled "Introducing OpenAI Frontier," positions the platform around a specific thesis: what slows enterprise AI adoption is not model intelligence but how agents are built and run inside organizations. OpenAI says it has worked with over 1 million businesses over the past few years and argues the gap between what models can do and what teams can deploy keeps widening as AI improves.

"At OpenAI alone, something new ships roughly every three days, and that pace is getting faster," the company writes.

The numbers OpenAI is citing

OpenAI claims 75% of enterprise workers say AI helped them do tasks they could not do before. It also cites three deployments, described without naming the companies:

• A major manufacturer reduced production optimization work from six weeks to one day using agents.
• A global investment company deployed agents end-to-end across its sales process, opening up over 90% more time for salespeople to spend with customers.
• A large energy producer increased output by up to 5% with agents, which OpenAI says adds over a billion dollars in additional revenue.

The company also details an internal-style deployment at a hardware company: millions of hardware tests failed, and engineers spent nearly half their time manually hunting root causes through logs, docs, and code. With AI coworkers pulling together simulation logs, internal documentation, workflows, and code to run end-to-end investigations, root-cause identification dropped from roughly 4 hours per failure to a few minutes, saving thousands of engineering hours annually.

State Farm signs on

State Farm is the flagship named customer. Joe Park, the insurer's Executive Vice President and Chief Digital Information Officer, said in a statement:

"Partnering with OpenAI helps us give thousands of State Farm agents and employees better tools to serve our customers. By pairing OpenAI's Frontier platform and deployment expertise with our people, we're accelerating our AI capabilities and finding new ways to help millions plan ahead, protect what matters most, and recover faster when the unexpected happens."

For OpenAI, a named Fortune 100 insurer matters commercially. Enterprise contracts of that scale anchor the revenue base the company needs, and Frontier gives it a packaged answer to competing enterprise agent platforms from Google, Microsoft, and Anthropic, though OpenAI does not name competitors in the post.

What Frontier actually does

OpenAI frames the platform around four capabilities, analogized to what human employees need to succeed: shared context, onboarding, hands-on learning with feedback, and clear permissions.

Understand the work. Frontier connects siloed data warehouses, CRM systems, ticketing tools, and internal applications into a shared business context. OpenAI describes this as a semantic layer for the enterprise that all AI coworkers can reference to operate and communicate effectively.

Plan, act, and solve problems. Technical and non-technical teams can deploy AI coworkers that handle tasks people already do on a computer: reasoning over data, working with files, running code, and using tools in an agent execution environment. Agents build memories from past interactions to improve performance over time. They can run across local environments, enterprise cloud infrastructure, and OpenAI-hosted runtimes, and Frontier prioritizes low-latency access to OpenAI's models for time-sensitive work.

Improve quality on real work. Built-in evaluation and optimization tooling shows human managers and AI coworkers what is working, so good behaviors improve over time. "This is how agents move from impressive demos to dependable teammates," the post reads.

Identity, permissions, and boundaries. Each AI coworker gets its own identity with explicit permissions and guardrails, which OpenAI says makes the agents usable in sensitive and regulated environments. Enterprise security and governance are built in.

No replatforming required

A central selling point is that Frontier works with existing systems rather than replacing them. Enterprises connect existing data and applications where they live, using open standards, with no new formats and no abandonment of agents or applications already deployed. OpenAI argues the payoff is that AI coworkers are accessible through any interface, whether that is ChatGPT, workflows with Atlas, or existing business applications, and whether agents were built in-house, acquired from OpenAI, or integrated from other vendors.

The open-standards posture also extends to third-party developers. Because agent applications often fail from lack of context, scattered data, and one-off integrations, Frontier lets software teams plug in and build agents that draw on the same shared business context, with controls, from day one.

OpenAI is seeding that ecosystem with a small group of Frontier Partners: Abridge, Clay, Ambience, Decagon, Harvey, and Sierra. These AI-native builders will work closely with OpenAI to learn what customers need, design solutions, and support deployment. OpenAI says it will expand the program over time to more builders focused on enterprise AI.

Humans in the loop, by design

Frontier is not software alone. OpenAI pairs customers with Forward Deployed Engineers (FDEs) who work side by side with enterprise teams to develop best practices for building and running agents in production. The FDEs also provide a direct connection to OpenAI Research, creating what the company describes as a feedback loop from business problem to deployment to research and back.

"As you deploy agents, we learn not just how to improve your systems around the model. We also learn how the models themselves need to evolve to be more useful for your work," OpenAI writes.

The model echoes Palantir's forward-deployed engineer playbook, a proven approach for landing high-value enterprise contracts in complex, regulated environments.

Why it matters

OpenAI is making a deliberate land grab for enterprise agent infrastructure at a moment when the category is unsettled. Its diagnosis is that fragmentation, not intelligence, is the bottleneck: agents deployed everywhere, each isolated in what it can see and do, each new one adding complexity instead of value. If Frontier's semantic layer and identity model become the connective tissue for enterprise agents, OpenAI captures a strategic position above individual applications, and its partner network locks AI-native vendors into its stack.

The stakes are also competitive in timing. OpenAI frames the story as a widening gap between early AI leaders and everyone else, and it is betting enterprises will move to close that gap on its platform rather than a rival's. The limited initial availability and the FDE-heavy delivery model suggest a deliberate scarcity strategy aimed at large, complex deployments first.

The next signal to watch is broader availability over the coming months, and whether OpenAI names additional customers beyond State Farm with measurable outcomes attached.